From 8a1e491ead09384be18affdb133dcd23440ba1db Mon Sep 17 00:00:00 2001
From: Staroom <31646341+eastmoe@users.noreply.github.com>
Date: Sat, 23 May 2026 17:00:23 +0800
Subject: [PATCH] =?UTF-8?q?=E7=BB=A7=E7=BB=AD=E6=94=BE=E5=AE=BDtorch\torch?=
 =?UTF-8?q?audio\omegaconf\k-diffusion\einops-exts=E7=9A=84=E7=89=88?=
 =?UTF-8?q?=E6=9C=AC=E9=99=90=E5=88=B6?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

这次新增改动：
requirements.txt (line 16)：torch>=2.6.0,<2.12
torchaudio>=2.6.0,<2.12
原因是当前 Torch 已到 2.12.0，但 Torchaudio 只有 2.11.0，放到 <3.0 可能装出二进制不匹配组合。

requirements_nodeps.txt (line 1)：同步放宽 omegaconf
修正 antlr4-python3-runtime>=4.9,<4.10，匹配 OmegaConf 2.3 的上游要求。

tokenization_llama.py (line 29)：给 Transformers 5.x 的 tokenizer 入口变化加 fallback：旧：transformers.tokenization_utils
新：transformers.tokenization_python / transformers.tokenization_utils_base

已核对的上游接口包括：
Transformers 4.37.2 和 5.9.0 的 SequenceSummary/GPT2SequenceSummary、Flash Attention helper、ONNX 移除、Llama tokenizer 入口、Hubert/Qwen2/Wav2Vec2Conformer 相关符号
Diffusers 0.37.1 的 randn_tensor、AdaLayerNormSingle、TimestepEmbedding
OmegaConf 2.3.0 的 antlr4 依赖要求
Torch/Torchaudio 当前可解析版本组合
---
 .../codeclm/models/llama/tokenization_llama.py         |  6 +++++-
 songgeneration/requirements.txt                        | 10 +++++-----
 songgeneration/requirements_nodeps.txt                 |  4 ++--
 3 files changed, 12 insertions(+), 8 deletions(-)

diff --git a/songgeneration/codeclm/models/llama/tokenization_llama.py b/songgeneration/codeclm/models/llama/tokenization_llama.py
index bf105b7..d8a2a60 100644
--- a/songgeneration/codeclm/models/llama/tokenization_llama.py
+++ b/songgeneration/codeclm/models/llama/tokenization_llama.py
@@ -26,7 +26,11 @@ from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple
 import sentencepiece as spm
 
 from transformers.convert_slow_tokenizer import import_protobuf
-from transformers.tokenization_utils import AddedToken, PreTrainedTokenizer
+try:
+    from transformers.tokenization_utils import AddedToken, PreTrainedTokenizer
+except ImportError:
+    from transformers.tokenization_python import PreTrainedTokenizer
+    from transformers.tokenization_utils_base import AddedToken
 from transformers.utils import logging
 
 
diff --git a/songgeneration/requirements.txt b/songgeneration/requirements.txt
index 306b588..7b9070b 100644
--- a/songgeneration/requirements.txt
+++ b/songgeneration/requirements.txt
@@ -1,20 +1,20 @@
 alias-free-torch>=0.0.6
 diffusers>=0.37.1,<0.38
 einops>=0.8.1
-einops-exts==0.0.4
+einops-exts>=0.0.4
 huggingface-hub>=0.34.0,<2.0
 julius>=0.2.7
-k-diffusion==0.1.1
+k-diffusion>=0.1.1,<0.2
 lameenc>=1.8.1
 librosa>=0.11.0
-omegaconf==2.2.0
+omegaconf>=2.2.0,<3.0
 openunmix>=1.3.0
 packaging>=24.0
 pydantic>=2.0
 PyYAML>=6.0
 safetensors>=0.4.3
-torch==2.6.0
-torchaudio==2.6.0
+torch>=2.6.0,<2.12
+torchaudio>=2.6.0,<2.12
 tqdm>=4.66.0
 transformers>=4.37.2,<6.0
 vector-quantize-pytorch>=1.22.17
diff --git a/songgeneration/requirements_nodeps.txt b/songgeneration/requirements_nodeps.txt
index 5fec19a..6505eac 100644
--- a/songgeneration/requirements_nodeps.txt
+++ b/songgeneration/requirements_nodeps.txt
@@ -1,2 +1,2 @@
-antlr4-python3-runtime==4.8
-omegaconf==2.2.0
+antlr4-python3-runtime>=4.9,<4.10
+omegaconf>=2.2.0,<3.0